Multichannel Acoustic Feature Enhancement Method for Rotating Machinery with Polynomial Eigenvalue Decomposition and Improved Morphological Filtering

As a non-contact monitoring technology, microphone arrays effectively address the issues of installation limitations or restricted flexibility of vibration sensors in certain scenarios. However, as a commonly used array signal processing technique, beamforming typically relies on accurate knowledge of the array geometry. When such information is inaccurate or unavailable, its performance may degrade significantly. To overcome this limitation, this paper proposes a novel fault diagnosis method for microphone array signals with polynomial eigenvalue decomposition (PEVD) and improved morphological filtering (MF). This method enables accurate bearing fault diagnosis without requiring prior information about the array geometry. First, PEVD is employed to capture the spatio-temporal correlations of multi-channel signals, enabling effective spatial filtering. A novel metric, the Sparse Energy Ratio (SER), is introduced to select the optimal signal component. Second, by incorporating MF, the Morphological Hat Product Difference Operator (MHPDO) is utilized to effectively extract the periodic impulse characteristics of the signal. Finally, envelope analysis is performed for feature extraction and fault diagnosis. The effectiveness of the proposed method is validated through both simulation and experimental data. The results demonstrate that the proposed method accurately identifies bearing faults under steady-state noise as well as in single and multiple interference noise environments, effectively enhancing the accuracy and robustness of fault diagnosis.